MATLAB: Does the output of IIR Band Pass filter have a huge spike when filtering a noisy sinusoid in Filter Design Toolbox 4.5 (R2009a)

Filter Design Toolbox

I am generating a noisy sinusoid using the SIN and WGN functions. Next I design a band pass IIR Butterworth filter using FDATOOL. I export the SOS and G variables to workspace. I convert these to a transfer function using SOS2TF and apply this filter on the signal using the FILTER command.

I see the output has a huge spike. I am wondering what went wrong.

Best Answer

The unexpected output is due to the quantization errors occuring during conversion of a SOS filter into a transfer function.

To resolve this issue export the filter as an object. Apply this filter object on the input signal using the FILTER command. This way filter does not have to be converted to TF and therefore generates expected results.

Related Solutions

MATLAB: Fir and iir filtering

You have to know your filter specifications. If you do, you can do both with fdesign.lowpass and fdesign.highpass

fdesign.lowpass and fdesign.highpass have different specification strings which support different filter designs.

The default string 'Fp,Fst,Ap,Ast' supports both FIR and IIR designs.

Here is an example:

Design a lowpass filter with a passband frequency of 1 kHz, a stopband frequency of 1.5 kHz, passband ripple of 0.5 dB, and stopband attenuation of 40 dB. The sampling frequency is 20 kHz

   d = fdesign.lowpass('Fp,Fst,Ap,Ast',1e3,1.5e3,0.5,40,20e3);
   designmethods(d)  %shows which designs are valid for this spec
   Hdeq = design(d,'equiripple'); % FIR filter
   Hbutt = design(d,'butter'); %IIR design

MATLAB: Has the algorithm for fdesign changed in Signal Processing Toolbox 6.13 (R2010a) as compared to the previous versions

This is expected behavior. In Signal Processing Toolbox 6.11 (R2009a) , there was an improvement in the design of minimum order equiripple lowpass filters which involved measuring the filter response to ensure that the filter order meets the design specifications. This is causing the increment in the filter order in R2009a and later versions.

As a workaround, design the filter with non-minimum order specifications thus forcing the same number of coefficients as in R2007b. However, this changes the filter specification and will return different values for the coefficients.

Best Answer

Related Solutions

MATLAB: Fir and iir filtering

MATLAB: Has the algorithm for fdesign changed in Signal Processing Toolbox 6.13 (R2010a) as compared to the previous versions

Related Question